Databases Reference
In-Depth Information
Narrow/Narrow
Narrow/narrow workloads are queries generated by standard reports and run execution on a narrow
set of columns, summarized tables, or aggregate structures returning a narrow set of columns in the
result set. The resulting complexity from these queries are impacted by the:
●
Length of the table
●
Distribution of the table across the storage
●
Data model
●
Time dimension
●
Semantic layer functionality
Typically, these queries exhibit the following traits:
●
Smaller I/O
●
High CPU consumption
●
High memory utilization
●
Multiple cycles of read
●
Heavy computes
Unstructured/semi-structured data
Another big impact on query performance comes from the integration of data in the query from
a non-RDBMS structure like Excel files, CSV files, and other storage structures. The impact on the
workload performance is the actual loading of the data into memory from the file for each cycle of
operation and managing the intermediate data sets. The queries affected here are across all the differ-
ent types of classifications discussed in this section, and they exhibit the following traits:
●
Large I/O cycles
●
High CPU consumption
●
High memory utilization
●
Heavy computes
●
Long waits of execution (from a perspective of time)
●
Lots of temporary storage
File-based data in the traditional data management platform is a threat to increasing and skewing
workloads since they are managed in a serial fashion in terms of reads and writes. In the traditional
data integration architecture, there is no parallelism involved in file-based operations due to consistency
management, which has been a big deterrent in the performance and negatively impacted workloads.
While queries by themselves are a problem when it comes to workload management, the associ-
ated data integration challenges that are present in this layer of the architecture make it one of the
most complex areas to design, maintain, and manage on an ongoing basis.
ETL and CDC workloads
Figure 8.4
shows the categorization of ETL and CDC workloads. Depending on the data type, the
associated business rule for transformation, and the underlying data model and architecture, the
